AUTHOR=Zheng Zhenhui , Xiong Juntao , Lin Huan , Han Yonglin , Sun Baoxia , Xie Zhiming , Yang Zhengang , Wang Chenglin 

TITLE=A Method of Green Citrus Detection in Natural Environments Using a Deep Convolutional Neural Network

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 12 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2021.705737

DOI=10.3389/fpls.2021.705737

ISSN=1664-462X

ABSTRACT=Accurate detection of green citrus in natural environment is a key step to realize the intelligent harvesting of citrus by robot. At present, the visual detection algorithms for green citrus in natural environment still have poor accuracy and robustness due to the color similarity between fruits and background. This study proposed a multi-scale convolutional neural network (CNN) named YOLO BP to detect green citrus in natural environments. Firstly, the backbone network, CSPDarknet53 was trimmed to extract high-quality features and improve the real-time performance of the network. Then, by removing redundant nodes of Path Aggregation Network (PANet) and adding additional connections, a bi-directional feature pyramid network (Bi-PANet) was proposed to efficiently fuse multi-layer features. Finally, three groups of green citrus detection experiments were designed to evaluate the network performance. The results showed that accuracy, recall, mean average precision (mAP) and detection speed of YOLO BP were 86.00%, 91.00%, 91.55% and 18 Frames Per Second (FPS), respectively, which were 2%, 7%, 4.30% and 1FPS higher than those of the YOLO v4. The proposed detection algorithm has strong robustness and high accuracy in the complex orchard environment, which provides technical support for green fruit detection in natural environments.